[snemvalts]

"I'd like a car that goes 300mph and gets 100mpg while doing it. I'm aware of a car that gets 100mpg but it is extremely slow."

You are describing fundamental tradeoffs. Getting more performance relative to model size and training token amount is what all of the labs are solving.

[docheinestages]

Labs are focusing on creating models, small or large, that perform well on various benchmarks, including general knowledge, domain-specific expertise, and agentic capabilities.

Asking for such a model while wanting to be small and fast would align with what you're describing, which I believe is different from what I'm pointing to.

The model I'm describing sacrifices domain knowledge and expertise for agentic reasoning and tool-calling capabilities at a reasonable speed.

Think of Cactus Compute's Needle [1].

[1] https://cactuscompute.com/blog/needle

[tossandthrow]

The core intuition here is probably that you can not separate reasoning and domain knowledge.

How are you going to use tools, if you don't have the context to use them?

Imaging if you could only think in terms of lambda calculus, and was asked to check the weather to let me know if the weekend is good for a hike.

[DennisP]

I think the idea here would be to not use your super smart but specialized model to check the weather. It's not obvious that it's impossible to (eg) remove most of its biology knowledge, without removing much of its ability to develop software (for example). If you're developing biology software, then don't use that particular compressed model.

(If you're claiming that it is impossible, and you have references you can share, then I'm honestly interested.)

[tossandthrow]

I don't think that is true.

I think more of the capabilities comes from the cross domains knowledge.

Anyways, software is also designed for a domain. The reason why it is so adept at making a fitness tracker is likely because it knows about biology.

[DennisP]

Sure, that's a plausible theory but I haven't seen that anyone's proved it.

An alternate theory is that models need lots of input knowledge to learn complex reasoning, but don't need so much at inference time. An example is arithmetic. Early in their training, models do arithmetic (poorly) by pattern-matching on memorized examples. Eventually they grok arithmetic and stop pattern matching, and then they don't use the examples anymore.

The fitness tracker doesn't take much knowledge of biology. Large models have quite a lot of biology knowledge that most people will never use. Same goes for lots of other topics. For basic knowledge that easily fits in context it could search the internet, or a local collection of introductory textbooks.